Multiparametric MRI-based biomarkers in the non-fluent and semantic variants of primary progressive aphasia

Abstract

Background:

The non-fluent (nfPPA) and semantic (svPPA) variants of primary progressive aphasia exhibit distinct clinical features. We investigated whether diffusion tensor imaging (DTI) and atlas-based volumetry (ABV) could reveal divergent patterns of longitudinal changes in brain white matter microstructure and gray matter volumes. Methods:

MRI datasets from 29 nfPPA, 27 svPPA, and 39 controls were analyzed. White matter fractional anisotropy (FA) differences were assessed using unbiased Whole Brain-based Spatial Statistics (WBSS) and Tract-Wise Fractional Anisotropy Statistics (TFAS). Gray matter volumetric differences were calculated by Atlas-Based Volumetry (ABV). A subset of 10 nfPPA and 6 svPPA patients underwent longitudinal MRI at 12 months. FA maps were correlated with disease severity (FTLD-CDR sum of boxes). A random forest classifier validated tracts of interest (TOI) and structures of interest (SOIs) selection as a proof-of-concept. Results:

At baseline, nfPPA showed frontal, callosal, and temporal white matter degeneration, while the left inferior longitudinal fasciculus (ILF) was predominantly involved in svPPA. Longitudinally, nfPPA exhibited frontal, callosal, and posterior temporal progression, while svPPA showed localized antero-posterior ILF progression. ABV aligned with the DTI analyses, demonstrating volumetric reductions in the frontal lobe for nfPPA and in temporal lobe and subcortical limbic structures in svPPA. Sub-clusters of white matter damage progression correlated with worsening FTLD-CDR scores. Random forest analysis identified the most discriminative TOIs and SOIs. Conclusions:

Distinct degeneration patterns emerged across nfPPA and svPPA, supporting early differential diagnosis and correlating with disease worsening. These findings support the utility of combined DTI and ABV in tracking disease progression.